Strategy · May 20, 2026 · 6 min read
The AI ROI question: how to fund AI with confidence
73% of AI projects stall before production. The fix isn't better models — it's funding the right use cases with a number your CFO trusts.
Most AI programmes don't fail in the lab — they fail in the business case. The model works; the funding, sequencing and ownership don't. The result: 73% of initiatives stall before production (Gartner).
Lead with the P&L, not the model
Before a line of code, score every candidate use case on value and feasibility. Fund the high-value, high-feasibility wins first — and tie each to a number you already track: hours saved, cost avoided, revenue unlocked.
Three numbers every AI business case needs
- Projected ROI per dollar invested (benchmark: 3.7×, Microsoft/IDC).
- Payback period — when the initiative turns cash-positive.
- The KPI it moves — and the baseline you'll measure against.
If you can't name the KPI it moves, it isn't a use case — it's a science project.
Get the prioritisation right and AI stops being a cost centre and becomes a compounding growth engine.